qcom_scm_call_atomic1() can crash with a NULL pointer dereference at
qcom_scm_call_atomic1+0x30/0x48.
disassembly of qcom_scm_call_atomic1():
...
<0xc08d73b0 <+12>: ldr r3, [r12]
... (no instruction explicitly modifies r12)
0xc08d73cc <+40>: smc 0
... (no instruction explicitly modifies r12)
0xc08d73d4 <+48>: ldr r3, [r12] <- crashing instruction
...
Since the first ldr is successful, and since r12 isn't explicitly
modified by any instruction between the first and the second ldr,
it must have been modified by the smc call, which is ok,
since r12 is caller save according to the AAPCS.
Add r12 to the clobber list so that the compiler knows that the
callee potentially overwrites the value in r12.
Clobber descriptions may not in any way overlap with an input or
output operand.
Reviewed-by: Bjorn Andersson <[email protected]>
Reviewed-by: Stephen Boyd <[email protected]>
Signed-off-by: Niklas Cassel <[email protected]>
---
drivers/firmware/qcom_scm-32.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/firmware/qcom_scm-32.c b/drivers/firmware/qcom_scm-32.c
index dfbd894d5bb7..4e24e591ae74 100644
--- a/drivers/firmware/qcom_scm-32.c
+++ b/drivers/firmware/qcom_scm-32.c
@@ -147,7 +147,7 @@ static u32 smc(u32 cmd_addr)
"smc #0 @ switch to secure world\n"
: "=r" (r0)
: "r" (r0), "r" (r1), "r" (r2)
- : "r3");
+ : "r3", "r12");
} while (r0 == QCOM_SCM_INTERRUPTED);
return r0;
@@ -263,7 +263,7 @@ static s32 qcom_scm_call_atomic1(u32 svc, u32 cmd, u32 arg1)
"smc #0 @ switch to secure world\n"
: "=r" (r0)
: "r" (r0), "r" (r1), "r" (r2)
- : "r3");
+ : "r3", "r12");
return r0;
}
@@ -298,7 +298,7 @@ static s32 qcom_scm_call_atomic2(u32 svc, u32 cmd, u32 arg1, u32 arg2)
"smc #0 @ switch to secure world\n"
: "=r" (r0)
: "r" (r0), "r" (r1), "r" (r2), "r" (r3)
- );
+ : "r12");
return r0;
}
@@ -328,7 +328,7 @@ u32 qcom_scm_get_version(void)
"smc #0 @ switch to secure world\n"
: "=r" (r0), "=r" (r1)
: "r" (r0), "r" (r1)
- : "r2", "r3");
+ : "r2", "r3", "r12");
} while (r0 == QCOM_SCM_INTERRUPTED);
version = r1;
--
2.14.3
On Thu 26 Apr 04:55 PDT 2018, Niklas Cassel wrote:
> qcom_scm_call_atomic1() can crash with a NULL pointer dereference at
> qcom_scm_call_atomic1+0x30/0x48.
>
Hi Niklas,
The change has been picked up for v4.18, as you can see here:
https://git.kernel.org/pub/scm/linux/kernel/git/agross/linux.git/log/?h=for-next
Regards,
Bjorn
On Thu, Apr 26, 2018 at 11:01:30AM -0700, Bjorn Andersson wrote:
> On Thu 26 Apr 04:55 PDT 2018, Niklas Cassel wrote:
>
> > qcom_scm_call_atomic1() can crash with a NULL pointer dereference at
> > qcom_scm_call_atomic1+0x30/0x48.
> >
>
> Hi Niklas,
>
> The change has been picked up for v4.18, as you can see here:
> https://git.kernel.org/pub/scm/linux/kernel/git/agross/linux.git/log/?h=for-next
Hello Bj?rn,
I just looked at today's linux-next tag: next-20180426.
I didn't bother to check Andy's tree.
Andy's for-next branch is included in linux-next, so it
will probably be included in tomorrow's linux-next tag.
Sorry for the noise.
Kind regards,
Niklas
On Thu, Apr 26, 2018 at 09:25:36PM +0200, Niklas Cassel wrote:
> On Thu, Apr 26, 2018 at 11:01:30AM -0700, Bjorn Andersson wrote:
> > On Thu 26 Apr 04:55 PDT 2018, Niklas Cassel wrote:
> >
> > > qcom_scm_call_atomic1() can crash with a NULL pointer dereference at
> > > qcom_scm_call_atomic1+0x30/0x48.
> > >
> >
> > Hi Niklas,
> >
> > The change has been picked up for v4.18, as you can see here:
> > https://git.kernel.org/pub/scm/linux/kernel/git/agross/linux.git/log/?h=for-next
>
> Hello Bj?rn,
>
> I just looked at today's linux-next tag: next-20180426.
> I didn't bother to check Andy's tree.
>
> Andy's for-next branch is included in linux-next, so it
> will probably be included in tomorrow's linux-next tag.
>
> Sorry for the noise.
I'm pretty bad about broadcasting which patches i am taking. Sorry bout that.
Andy